It looks like that code was never actually MIT licenced but had the licence-file in there by mistake. Then Roman made the repo public after the faulty licence got removed, because he (wrongly) believed the code actually was MIT-licenced before the change.
That's how I read it. Is it true? No idea but it might be.